Assessments of surface water quality through the use of multivariate statistical techniques: A case study for the watershed of the Yuqiao Reservoir, China

In light of the fact that water quality has been threatened by human activities, apportionments of potential pollution sources are essential for water pollution control. Multivariate methods were used to assess the water quality in the Yuqiao Reservoir and its surrounding rivers in northern China to identify potential pollution sources and quantify their apportionment. Fifteen variables at 10 sites were surveyed monthly in 2015-2016. The quality at this location was acceptable according to the water quality index (WQI), except for special parameters including chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and chlorophyll (chl alpha). Cluster analysis (CA) grouped these datasets into three seasonal groups, July-September, December-March, and the remaining months. Principal component analysis/factor analysis (PCA/FA) identified seven factors that accounted for 79.7%-86.4% of the total variance, and the main sources included cities, rural districts, industries, weather, fertilizers, upstream areas, and vehicles. Absolute principal component scores and multiple linear regression (APCS-MLR) modeling results show that the hierarchical contribution of main pollution sources was ranked in the following order: upstream (26.6%) > urban district pollution source (21.5%) > vehicle emission pollution source (10.9%) in the flood season, upstream (22.3%) > rural district pollution (19.8%) > fertilizer erosion (15.8%) in the normal season, and upstream (26.4%) > urban district pollution (19.0%) > fertilizer erosion (18.8%) in the dry season. Sources from upstream and urban districts explained the most proportion. The matrix was also subjected to positive matrix factorization (PMF). A comparison of PMF and APCS-MLR results showed significant differences in the contribution of potential pollution sources. The APCS-MLR model performed better, as evidenced by a more robust R (2) test. Measures should be discussed and implemented in managing upstream areas, sewage treatment facilities, and fertilizer and industrial application.